arXiv subjects

Chunming Wang

Publications and source records attributed to Chunming Wang.

6 recordsLinked to original sources

arXiv2026

ClinicalMC: A Benchmark for Multi-Course Clinical Decision-Making with Large Language Models

Large language models (LLMs) have been widely adopted in healthcare, yet they still encounter significant challenges in complex clinical decision-making scenarios. Existing benchmarks primarily assess LLM performance in single-course settings and lack systematic evaluation in multi-course scenarios, where a patient's condition evolves over time. To address this gap, we propose ClinicalMC, a benchmark for multi-course clinical decision-making. It includes 1,275 Chinese and 5,804 English samples across four stages from admission to discharge. These stages cover triage, first-course examination/diagnosis/treatment, subsequent multi-course examination/assessment/treatment, and final diagnosis. In ClinicalMC, patients in the English dataset undergo an average of 5.11 clinical courses, whereas those in the Chinese dataset undergo 3.42. To assess LLM performance, we construct a multi-agent evaluation framework that includes patient, examiner, and doctor agents. Based on the benchmark and framework, we design two experimental settings -- a single-turn static setting and a multi-turn dynamic setting -- and assess three categories of LLMs: 1) closed-source LLMs like GPT5-mini; 2) open-source LLMs like DeepSeek-V3.2; and 3) medical LLMs like HuatuoGPT-o1. Through extensive evaluation, we aim to better understand LLM performance in the medical domain and support its effective deployment in healthcare.

cs.AI↗

arXiv2026

Wetting-coupled phase separation as an energetic mechanism for active bacterial adhesion

The rapid adhesion of motile bacteria from dilute suspensions poses a fundamental non-equilibrium problem: hydrodynamic interactions bias bacterial motion near surfaces without generating stable confinement, while electrostatic interactions are predominantly repulsive. Here, combining experiments on Pseudomonas aeruginosa and Staphylococcus aureus in a polyethylene glycol/dextran aqueous two-phase system with large-scale hydrodynamic simulations, we identify wetting-coupled liquid--liquid phase separation (LLPS) as an energetic trapping mechanism for bacterial adhesion. When bacteria partition into a phase that preferentially wets the substrate, interfacial free-energy minimization creates a deep energetic trap that stabilizes adhesion and induces lateral clustering via capillary interactions. Crucially, bacterial motility plays a dual role: at low phase volume fractions, activity enhances transport into the wetting layer and promotes accumulation, whereas at higher phase volumes it suppresses adhesion through the formation of self-spinning droplets that generate hydrodynamic lift opposing interfacial trapping. Our results establish wetting-coupled LLPS as a generic physical route governing interfacial organization in active suspensions. This provides a unified energetic framework for bacterial adhesion in complex fluids, with broad implications for deciphering bacterial-cell interactions and controlling biofilm formation.

cond-mat.soft↗

arXiv2025

Active motility and wetting cooperatively regulate liquid-liquid phase separation

Liquid--liquid phase separation in aqueous two-phase systems is fundamental across physical and biological sciences. While well understood for passive mixtures, how it is regulated by active agents such as motile bacteria remains largely unexplored. By combining experiments on Pseudomonas aeruginosa in a dextran--polyethylene glycol mixture with hydrodynamic simulations, we show that the coupling between bacterial activity and interfacial wetting converts self-propulsion into mechanically effective interfacial stresses, giving rise to a robust sequence of morphologies, including self-spinning droplets, elongated droplet chains, and branched capillary-like clusters. More importantly, it gives activity a dual kinetic role: activity suppresses coarsening in the droplet regime through rotation-induced hydrodynamic repulsion, but accelerates coarsening when dextran is the minority phase, where wetting-mediated attraction drives aggregation. To probe the biological relevance of this mechanism, we further show that a biofilm-associated protein can act as an interfacial ``wetting glue,'' promoting bacterial clustering even in dilute suspensions. Our findings establish activity--wetting coupling, rather than activity or wetting alone, as a mechanism by which active motility regulates pattern morphology and coarsening dynamics, and reveal a physical route to enhancing bacterial aggregation from dilute suspensions.

cond-mat.soft↗

arXiv2022

Anomalous magneto-thermoelectric behavior in massive Dirac materials

Extensive studies of electron transport in Dirac materials have shown positive magneto-resistance (MR) and positive magneto-thermopower (MTP) in a magnetic field perpendicular to the excitation current or thermal gradient. In contrast, measurements of electron transport often show a negative longitudinal MR and negative MTP for a magnetic field oriented along the excitation current or thermal gradient; this is attributed to the chiral anomaly in Dirac materials. Here, we report a very different magneto-thermoelectric transport behavior in the massive Dirac material ZrTe5. Although thin flakes show a commonly observed positive MR in a perpendicular magnetic field, distinct from other Dirac materials, we observe a sharp negative MTP. In a parallel magnetic field, we still observe a negative longitudinal MR, however, a remarkable positive MTP is observed for the fields parallel to the thermal gradients. Our theoretical calculations suggest that this anomalous magneto-thermoelectric behavior can be attributed to the screened Coulomb scattering. This work demonstrates the significance of impurity scattering in the electron transport of topological materials and provides deep insight into the novel magneto-transport phenomena in Dirac materials.

cond-mat.mtrl-sci↗

arXiv2018

Scaling behavior of dynamic hysteresis in Na0.5Bi4.5Ti4O15 bulk ceramics

The ferroelectric hysteresis loops of sodium bismuth titanate Na0.5Bi4.5Ti4O15 bulk ceramics were measured under periodical electric field in range of frequency from 0.01Hz to 100Hz and field from 10kV/cm to 150kV/cm. The three-stage scaling behavior of dynamic hysteresis was investigated in Na0.5Bi4.5Ti4O15 bulk ceramics. The scaling behavior at low amplitude of electric field is described as is proportional to f^(-0.122)E0^3.30 for low frequency and is proportional to f^(-0.122)E0^3.15 for high frequency. , f and E0 represent the area of hysteresis loop, frequency and amplitude of periodic electric field, respectively. At E0 around coercive field, scaling behavior takes the form of is proportional to f^(-0.11)E0^4.28 for low frequency and is proportional to f^(-0.11)E0^4.17 for high frequency. At high E0, we obtained is proportional to f^(-0.04)E0^2.90 for low frequency and is proportional to f^(-0.06)E0^2.75 for high frequency. The contribution to scaling relation mainly results from reversible of ferroelectric domain switching at low E0, the velocity of domain wall motion at E0 around coercive field and simultaneously reversible and irreversible domain switching at high E0.

cond-mat.mtrl-sci↗

arXiv2017

Accelerating SGD for Distributed Deep-Learning Using Approximated Hessian Matrix

We introduce a novel method to compute a rank $m$ approximation of the inverse of the Hessian matrix in the distributed regime. By leveraging the differences in gradients and parameters of multiple Workers, we are able to efficiently implement a distributed approximation of the Newton-Raphson method. We also present preliminary results which underline advantages and challenges of second-order methods for large stochastic optimization problems. In particular, our work suggests that novel strategies for combining gradients provide further information on the loss surface.

cs.LG↗